Datatype meeting

14 Jan 2021

Attendees

• Guiomar Niso

• Franco Pestilli

• Soichi Hayasi

• Giulia Berto

• Aurore Bussalb (Unlicensed)

Discussion on MEG and EEG input formats

• EEG datatype in Brainlife: https://brainlife.io/datatype/5ddf1381936ca39318c5f045

• MEG datatype in Brainlife: https://brainlife.io/datatype/5dcecaffc4ae284155298383 (everything is marked “optional” because we can’t predict if a .fif file or a .ctf file will be used, but a datatype tag “fif” or “ctf” is added in addition to the datatype MEG)

Brainstorm reads all the MEG and EEG formats (no need to convert anything).
MNE Python does as well read a lot of formats. Same for FieldTrip (FT).

It’s important that the datatype stay flexible so that the App will be used by many.

Create Apps that can read most common formats (Apps don’t have to support all formats) and create converting Apps to famous datatypes (for instance in BIDS, EEG formats are limited because a lot of converters exist). But keep in mind that you need space to store converted data.

→ Initially, better to only work with one format then create converting Apps.

→ Start to write Apps reading only .fif formats? (To be discussed with Maximilien CHAUMON , Laurent HUGUEVILLE and GEORGE Nathalie) Then maybe add .ctf? Great if the App could read both .ctf and .fif files.

Answer fromGEORGE Nathalie : I think it would be good to read also a (native) EEG format; beware of not being stuck with MNE also: what does FT read and write?

Comment fromMaximilien CHAUMON : réservons un créneau avec Alex Gramfort dans quelques semaines pour lui poser toutes ces questions.

Comment from GEORGE Nathalie : please also consider the hdf5 format as possible format (https://fr.wikipedia.org/wiki/Hierarchical_Data_Format; https://www.neonscience.org/resources/learning-hub/tutorials/about-hdf5), it is already in Brainstorm (to check).

Outputs of Apps

What datatypes? If the code is in Matlab or Python, the datatype may change (to be discussed during another meeting).

Questions for next meetings

for datatype meeting

• in https://brainlife.io/datatype/6000737faacf9ee51fa691cb, do we add calibration files such as a fine_calibration_file.dat and a crosstalk_compensation_file.fif? These files are used in the function mne.preprocessing.maxwell_filter and in mne.preprocessing.find_bad_channels_maxwell.

for meeting with Alex

• How much does MNE rely on the fif file format. Does it make sense to have data readers in all formats, and handle everything in fif internally? Is there a special case for the ctf .ds format? Question asked to Richard Höchenberger: MNE relies on .fif internally.

• Head position alignment: before, after, independently of Maxfilter?

Discussion with Aina Puce following this meeting

25 Jan 2021

• Question of creating “data types” for different “data formats” (vendors) and different “types of data” (EEG, MEG, EOG… covariance matrix, calibration file, position file, freesurfer segmentation…)

Suggested “philosophy”:

1. Several import datatypes depending on vendors convert to toolbox-specific (MNE for now) dataformat

2. Processing datatypes with specific required files for each processing task

3. Several export datatypes

• A link to discussion on dataformat issues in BIDS working group: https://github.com/bids-standard/bep021/issues/1